viable/strict/1789203000: [inductor] Make manual wait-user repair linear (#195616)
- PyTorch: 838 events in the last 90 days
- PyTorch: 827th Release in the last 90 days
- Previous: earlier the same day · trunk/363267c09a72be003607d9e0e50df93150e4cb9a: [torchcomms hash update] update the pinned torchcomms hash (#196854)
What happened
Manual collective bucketing can replace wait outputs with graph nodes that appear later than their existing consumers. The previous repair moved each consumer separately and rebuilt the full node-position map twice per move. Long dependent chains therefore caused quadratic work and made large sharded graphs appear to hang. Run the stable topological sort once when a replacement introduces a backward data edge. Prese…
Summary assembled by rule from the sources below